Humans Take on AI Songwriting — but the Machine May Already Be Losing

Humans Take on AI Songwriting — but the Machine May Already Be Losing
The contest began with a deliberately provocative question: in an age of pervasive AI, is human creativity becoming a waste of time? New York Times reporter Matt Richtel assembled four lifelong musicians for a weekend studio session, joining them to write and record songs almost from scratch. Their opponent was an AI system asked to produce its own tracks.

The Times made listeners the final judges. It presented paired songs — including “Bury Me” and “Sweetgrass Clay” — and asked audiences to pick a favorite before revealing how each was made. The stated challenge was not merely whether humans could create the better song, but whether listeners could distinguish it from AI at all.

The human side of the exercise emphasized the messiness that machine generation bypasses: musicians standing together, listening, closing their eyes, arguing through choices and gradually shaping a song. Richtel described the setup as a “cage match: carbon life-forms versus silicon chips.”

But one account of the experiment saw little suspense in the result. It called the matchup “not much of a competition,” arguing that “the AI songs suck” and that, played side by side, the human and machine-made tracks are easy to tell apart.

That critique does not deny AI’s efficiency; it turns it into the point. The diary-like record of both workflows, it argues, exposes how low-effort AI music can be compared with the trial, disagreement and lived experience behind the human recordings.

For the Times, the listener poll leaves room for doubt and invites the public to weigh the evidence. For its critic, the experiment has already delivered its verdict: the machine may generate a song in moments, but it has not yet replicated the hard-won texture of people making one together.

https://foxvector.com